LED Detectives: what made these LEDs fail??

[php_44]

Hi All,

Though a seasoned EE, I'm not a quantum mechanic

, so I figured I'd pick some brains semiconductor brains. I'm looking for an LED forensic expert (since this might be a LED crime - possibly a coroner)

!!

I purchased a batch of InGaN based white LEDs that were returned to a vendor by an initial customer and were probably not retested by the vendor before being sold to me. Half the LEDs were good. Half would not light, and were failed as a low resistance short circuit when a low current was passed through them. I discovered that at higher currents some would light feebly. Further I discovered that a very brief high current pulse would mostly cure them - and allow them to work at close to expected brightness.

What sort of damage is there to the die to cause this behavior? Were they zapped by static electricity from poor handling? Were they subjected to reverse voltage breakdown?

The damage was cleared by a brief zap, however, I'm not going to keep the bad ones as I'm concerned about reduced life from the initial damage and damage from the apparent remedy. Would they live long and prosper if I "recovered" them?

I'm a little concerned because a few of the LEDs were failed "open" and had obviously burned plastic above the die - indicating big time over current --or perhaps an attempt to "recover" them by the vendor? I hope I have not been sold some repaired lemons.

I'm sending back the bad ones - but am still very curious. Any ideas?

 

[Marshall Johnson]

Well I'm not an expert either, but I'll take a shot at this.

I have heard people say before that they've fried LED's by overdriving them too much, and they were able to "fix" some of them by seriously overdriving them for a brief moment. I've never tried this myself, so I can't say for sure if that's what happened.

But since I can't think of anything else, this is my best guess: The original owner of the LED's burned them out by not limiting the current going to them. He probably didn't know what he was doing and hooked them up directly to some batteries without a current limiting resistor. After frying a bunch of them, he sent them back thinking they were defective.

That's just my 2 cents. Make that 1 cent because it's just a guess.

 

[php_44]

More points of interest. According to the vendor - these LEDs were returned for unacceptable color variation. I don't know if I believe the customer's reason. These had obviously been through an automated lead forming/cutting machine. The leads are exactly uniform. It is difficult to tell with certainty, but it is possible that they were unsoldered from a circuit board for return to the vendor. Many had some evidence of solder on the leads - but it's so uniform they must have been part of an automated assembly process. No flux residue, and very clean uniform solder. Most of the non-working LEDs had no solder evidence. This was a large lot which I got a small fraction of. Probably a many-thousand dollar argument between the original purchaser and the vendor when the return was settled. The lead forming/cutting rendered them unreturnable to normal vendor stock.

If the LEDs failed from being driven too hard, the original customer did it in a quantity of machine assembled boards that collectively used a huge number of the LEDs. That would be an embarrassing engineering mistake.

hmmmmm....

 

[Evan]

I've noted strange failures on blue & white LEDs, too. I've got a blue that will not light on 20ma of regulated DC current, will flash bright and then go dim on three used up AAAs in series, and lights just fine on the AC pulses from a brinkman flashlight circuit. The white also seemed completely dead on 20ma DC, but lit fine on the brinkman circuit, and then it was fine on DC, too! I know I tested it on DC 3 or 4 times, it was dead, but not anymore. Sounds like a similar phenomena.

 

[The_LED_Museum]

I've had them go in unusual ways too. One was part of a 3-series string, and failed "shorted". It emitted no light whatsoever, yet the current to the remaining two didn't increase significantly. And just when I had time to get in there and change it out, it came back on. So I left it in. It did this for a few months before I removed the installation entirely.

One common failure mode in GaN LEDs is damage to the quantum well structure induced by electrostatic discharge. This is usually detectable by a threshold test; the lamp's starting threshold voltage will be higher than normal, yet the LED will draw relatively high current from its power source well below threshold. When you return a "defective" LED, this is how Nichia and others can tell if you fried it or if it was actually defective from the start. The damage to the LED occurs at the atomic level, and is generally regarded as being irreparable. If enough of the quantum well structure remains unaffected, the LED may appear to work fine after being subject to various formed pulses of various amplitudes. I do *not* know the long-term outlook on devices abused in this manner.

Failures induced by excessive drive current are usually thermal in origin. Cratering of the ball bond, lifting of the bond wire from the ohmic contact (the metallization zone), and failure of the die attach can result from overdriving. Those LEDs that suddenly start flickering on and off at around 5-10Hz probably have a lifted bond wire; as the device heats up, the bond wire tears off the die causing it to go out; then as the parts cool, the connection is restored. Lather, rinse, repeat.

Other LED failure modes include:

Die attach migration.
This occurs when the material used to bond the die with the slug migrates up the side of the die over time, eventually lowering the device's resistance or shorting it out entirely. This type of failure occurs mainly with silver epoxide die attach materials; it happens much less frequently with modern eutetic die attachment processes.

Cratering.
This can be caused by lifting the bond wire off the die; which pulls up a chunk of die material off the emitter. Excessive current can also cause an area of high resistance and heating at the ball bond; eventually leading to a crack at the metallization zone and leading to the same type of breakage.

Purple Plague.
A failure that can be caused when gold and aluminum are used together inside an LED. At high drive currents where tempertures go way up, intermetallics begin to form at the interface between aluminum and gold. This process is called Kirkendahl voiding, and leads to brittle and broken interconnects that fail electrically because they fail mechanically. This intermetallic compound often has a purple color to it, hence the name.

Sudden expansion of encapsulant.
This is the failure mode that most LED experimenters are familar with most: the LED makes a loud "SNAP!" and then you smell this stinky smell in your room. Examination of the device shows catastrophic damage of varying degrees, ranging from the formation of a vent in the side of the encapsulant (where all the magic smoke escaped from) to total destruction of the device, where nothing but the metal leadframes attached to test clips remain. Severe overcurrent (15x-100000x normal value) causes this type of defect.

 

[star882]

The LEDs probably have a defect.
A whisker of metal shorted out the LED.
When you zapped it with a brief surge, the whisker blows like a fuse.